Research topic detection in scientific articles using a hybrid BERT integrated telescopic vector tree model with emperor penguin enhanced NSGA II optimization

• Published in: Scientific reports Vol. 15; no. 1; pp. 37201 - 18
• Main Authors: Krishnan, Keerthi, Easwarakumar, K. S.
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 24.10.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 639/705/117; 639/705/794; Accuracy; Algorithms; Artificial intelligence; Computer science; Data mining; Datasets; Deep learning; Humanities and Social Sciences; Hybrid BERT model; Keywords; Machine learning; Multi-objective optimization; multidisciplinary; Natural language processing; Neural networks; NSGA - II optimization; Optimization; R&D; Research & development; Research topic detection; Science; Science (multidisciplinary); Social networks; Telescopic vector -tree; Text analysis; Trees; Trends; NSGA - II optimization; Text analysis; Multi-objective optimization; Research topic detection; Telescopic vector -tree; Hybrid BERT model
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-21145-9

A vast number of research articles are published as a result of the enormous expansion in scientific study. An effective search from this collection of articles is made possible by using the correct search terms. Novice researchers find it challenging to search for articles that are pertinent to their areas of interest and to select the relevant keywords. Numerous studies suggest that optimization algorithms are effective in improving topic identification and detection. This provides researchers a valuable tool to help them traverse the deluge of research publications and keep abreast of the most recent developments in their field. By combining a TV-Tree (Telescopic Vector Tree) and Hybrid BERT (Bidirectional Encoder Representations from Transformers) with EPO (Emperor penguin optimization) enhanced NSGA - II (Nondominated Sorting Genetic Algorithm-II), this paper suggests a novel method for identifying research topics. The Hybrid BERT model combines the strengths of BERT and other machine learning algorithms to improve topic detection accuracy, while the TV-Tree based system provides an efficient and scalable framework for organizing and retrieving research topics. This method seeks to attain state-of-the-art performance in research topic detection by utilizing the advantages of both approaches, allowing for more effective and efficient categorization and retrieval of research papers. In order to determine the most important research subjects across a range of disciplines, this study suggests a revolutionary topic identification technique. This system uses sophisticated algorithms to examine enormous volumes of data in order to identify areas of interest and trending themes. Experimental results demonstrate the effectiveness of this approach in detecting prominent research topics, leading to potential accuracy improvements and significantly outperforming baseline models. This provides valuable insights for researchers, academics, and industry professionals seeking to stay updated on the latest developments in their fields.